[PATCH] malloc: Improve performance of __libc_malloc

[Wilco Dijkstra]

Hi Cupertino,

> Just wondering why you choose to set __attribute_noinline__ to
> __libc_malloc2 ?
> I know it is tailcalled, but having a single caller, would it not make
> sense to inline?

The tailcall means it ends up being a leaf function and thus doesn't
need a frame. If you inline it, you execute a prolog and epilog even
if you just use the tcache fastpath.

Cheers,
Wilco